04 January 2011

Signals of Anthropogenic Climate Change in Disaster Data

[UPDATE: Joe Romm blows a gasket on this paper.  Have a look at his post and you'll get a sense why it is that he dodged an opportunity to debate me last year (an offer that remains open) even after a donor promised $20,000 to Doctors Without Borders if he'd participate.] 

At ClimateWire, Evan Lehmann has a lengthy overview (also here at the NYT) of a new paper by Ryan Crompton, John McAneney (both of Macquarie University) and me on detecting signals of human-caused climate change in disaster losses. Ryan, who is wicked smart, did the heavy lifting and heavy thinking on it and deserves credit for what should be a widely influential article.

Lehmann writes in the article:
Economic losses are seen as a potent storyteller about climate change. If greenhouse gases could be shown to increase financial damages, that might accelerate efforts to develop stronger buildings codes, influence insurance prices for coastal homes, and discourage development in risk-prone areas.

But the research tells a different story, at least for hurricanes. As a backdrop, it uses a landmark study published in Science last January finding that the number of strongest hurricanes, categories 4 and 5, could double in 100 years because of climate change.

The researchers begin by assuming that's true. Then they apply hurricane damage data from the past century to those future hazards, adjusting for growth in population, inflation and wealth.

The results indicate that future hurricane damages won't produce a tangible "climate signal" for at least 120 years, and perhaps not for 550 years. The average time before a signal might be seen is 260 years, according to the combined findings of an 18-model ensemble used by the researchers. In that year, 2271, climate change is expected to increase damage by 106 percent, more than double.

The researchers know this is a touchy topic. It could be perceived as an effort to downplay the impacts of climate change, or be seen with alarm by environmentalists advocating for action now to cut carbon pollution.

"It's not to dispute that [global warming] is happening or what influence it will have on hurricanes," said Ryan Crompton, a co-author and a catastrophe risk expert with Risk Frontiers, a research organization at Macquarie University near Sydney, Australia, that is funded in part by the insurance industry.
The study that we build off of is Bender et al. 2010 which argues that under their projected changes in the behavior of hurricanes a signal will not be detectable in the geophysical data until later this century.  If it takes that long to detect a signal in the geophysical data, it is just common sense that it will take longer to see that signal emerge in loss data.  Tom Knutson, a co-author of Bender et al. agrees:
. . . the economic loss research confirms what Tom Knutson, a research meteorologist with the National Oceanic and Atmospheric Administration and an expert on climate change and hurricanes, has considered common sense: Climate change will reveal itself first in the hazard, then in the damage.

"Nothing too earth-shattering here," Knutson said. "This study actually doesn't tell me much that I already sort of have a notion of anyway. Namely, that if one is looking for a signal of climate change in the hurricane record, damage is not the first place to go to look for it."
Our 2006 Hohenkammer workshop with Munich Re reached this same conclusion. Our paper makes a pretty convincing and straightforward argument (in my view) and should make it clear why it is just plain wrong to attribute recent disasters (and even recent trends in disasters) to human-caused climate change.

While our paper focused on Atlantic hurricanes, the same sort of dynamics are likely to be present for other phenomena in other locations, as well as in aggregated global loss data.  This paper provides a convincing reason why the literature review of normalization studies conducted by Bouwer last year did not find any peer reviewed papers detecting and attributing trends in disasters to greenhouse gas emissions.  Such attribution is not likely in the near term, at least based on the magnitude of changes projected by Bender et al.

The paper is forthcoming in Environmental Research Letters -- here is the citation, abstract and concluding section:
Crompton, R. P.,  R. A. Pielke Jr. and K. J. McAneney, 2011 (forthcoming). Emergence time scales for detection of anthropogenic climate change in US tropical cyclone loss data, Environmental Research Letters V. 6, No. 1.


Recent reviews have concluded that efforts to date have yet to detect or attribute an anthropogenic climate change influence on Atlantic tropical cyclone (of at least tropical storm strength) behaviour and concomitant damage. However, identification of such influence cannot be ruled out in the future. Using projections of future tropical cyclone activity from a recent prominent study we estimate the time it would take for anthropogenic signals to emerge in a time series of normalized US tropical cyclone losses. Depending on the global climate model(s) underpinning the projection, emergence time scales range between 120 and 550 years, reflecting a large uncertainty. It takes 260 years for an 18-model ensemble-based signal to emerge. Consequently, under the projections examined here, the detection or attribution of an anthropogenic signal in tropical cyclone loss data is extremely unlikely to occur over periods of several decades (and even longer). This caution extends more generally to global weather-related natural disaster losses.


This study has investigated the impact of the Bender et al [11] Atlantic storm projections on US tropical cyclone economic losses. The emergence time scale of these anthropogenic climate change signals in normalized losses was found to be between 120 and 550 years. The 18-model ensemble-based signal emerges in 260 years.

This result confirms the general agreement that it is far more efficient to seek to detect anthropogenic signals in geophysical data directly rather than in loss data [14]. It also has implications for the emergence time scale of anthropogenic signals in global weather-related natural disaster losses given these losses are highly correlated with US tropical cyclone losses (supplementary discussion and supplementary table 1). Our results suggest that the emergence time scales are likely to be even longer than those determined for US tropical cyclone losses given that different perils will have different sensitivities to future anthropogenic climate change and may even change in different directions. We note that US tropical cyclone losses may become increasingly less correlated with global weather-related records as the loss potentials of developing countries in particular continue to rise rapidly, irrespective of future changes in climate [15]. This means that the relationship between the signal emergence time in US tropical cyclone losses and global losses may weaken over time.

Based on the results from our emergence time scale analysis we urge extreme caution in attributing short term trends (i.e., over many decades and longer) in normalized US tropical cyclone losses to anthropogenic climate change. The same conclusion applies to global weather-related natural disaster losses at least in the near future. Not only is short term variability not ‘climate change’ (which the IPCC defines on time scales of 30 to 50 years or longer), but anthropogenic climate change signals are very unlikely to emerge in US tropical cyclone losses at time scales of less than a century under the projections examined here.

Our results argue very strongly against using abnormally large losses from individual Atlantic hurricanes or seasons as either evidence of anthropogenic climate change or to justify actions on greenhouse gas emissions. There are far better justifications for action on greenhouse gases. Policy making related to climate necessarily must occur under uncertainty and ignorance. Our analysis indicates that such conditions will persist on timescales longer than those of decision making, strengthening the case for expanding disaster risk reduction in climate adaptation policy [15].
If you'd like a pre-publication copy of the paper please send me an email at pielke@colorado.edu.


  1. Quote, "Climate change will reveal itself first in the hazard, then in the damage."

    Quote, "Based on the results from our emergence time scale analysis we urge extreme caution in attributing short term trends (i.e., over many decades and longer) in normalized US tropical cyclone losses to anthropogenic climate change. The same conclusion applies to global weather-related natural disaster losses at least in the near future. Not only is short term variability not ‘climate change’ (which the IPCC defines on time scales of 30 to 50 years or longer), but anthropogenic climate change signals are very unlikely to emerge in US tropical cyclone losses at time scales of less than a century under the projections examined here."

    Well that knocks the new mantra of "climate disruption" well and truly on the head.

    It appears that climate change has always been in the slow lane.

  2. Regards your update, Romm's Rant has its own:
    UPDATE: Contrary to what you may read on the blogosphere, no one has ever contact me with an offer to me to debate for charity, even many months after I noted that fact. In any case, I have explained why one doesn’t debate folks like Pielke (see “Debate the controversy!“). To paraphrase Juan Cole’s advice to climate scientists on how to avoid being Swift-boated, any debate or broadcast that pits a serial misinformer or misrepresenter against someone defending climate science is automatically a win for the misinformer, “since a false position is being given equal time and legitimacy.” That’s why 99% of the articles or blog posts you read by people demanding some climate science defender debate someone are by other serial misinformers.

    Wow - can't possibly give equal time and legitimacy to the other debater. Wow. The thing is, you offer him a stage to clearly state his case, offering him equal time and legitimacy, yet he is unable to reciprocate.

  3. It is discouraging to me that a blog of such solipsistic snarkitude could get all the excellent reviews on the right sidebar of the blog.

  4. Roger I assume this is the paper you alluded to a while back. looks interesting. I'm curious if you also looked at other basins. If not, would you expect the same kind of results?

    I admit that it seems a little counterintuitive that a doubling in hurricanes takes so long to show up in the damage record. Is it simply a matter of the record being so noisy?

  5. -4-Marlowe

    Thanks for the question. The answer is pretty simple -- the Bender et al. paper that we build off of finds that the detection of a signal in the actual physical climate data under the changes that they project would not occur until the latter part of the next century.

    We should not expect detection of a signal in damage data for a longer time period -- as Knutson says in the NYT interview, this is kind of obvious when you think about it (but you have to understand Bender et al. first -- ironically Romm's post has a figure from Bender et al. at the top, he obviously read neither paper;-).

    I would fully expect to see similar results in other basins. More on that at a later date.

    Thanks for the questions about the content of the paper, pretty rare in the blogosphere it seems;-)

  6. Well here's another one about the paper. I noted this sentence: "The same conclusion applies to global weather-related natural disaster losses at least in the near future."

    Storms are a rare pehnomenon, so it is natural that statistical efforts to find a signal will be slow due to small sample size and high volatility. That is why it is "far more efficient to seek to detect anthropogenic signals in geophysical data directly".

    My question relates to storms vs other types of extreme climate events. Most of the paper seems to be about storms, but there seem to be some brief statements that the same conclusions apply to all extreme weather events. Would you way that the conclusion equally applies to events like heat waves, droughts, and floods? And I'm not talking about economic losses, just trends in the occurrence becoming significant.

  7. -6-Dean

    Extreme events are, by definition, rare events, so detecting a signal of change is difficult in the best of circumstances. Where there is large variability on decadal and longer timescales the detection becomes even more difficult.

    Our paper is about economic losses, not the physical events themselves. The analysis Bender et al. 2010 did for projected hurricanes behavior (which was focused on the physical events themselves) argued that even under their projected changes in storm behavior, it would be the second half of the 21st century before detection would occur. I think you are on safe ground concluding that for other phenomena with similar variability and similar magnitudes of projected changes, the detection timescale will be similar. Detecting signals in economic losses will always be longer, for obvious reasons.

    The analysis is pretty straightforward so it can be easily applied to other phenomena. I wouldn't be surprised to see some phenomena with timescales shorted than we found for hurricane losses and others longer.

    It is just some math, which makes me surprised to see some of the reactions to the paper.

  8. As far as economic losses go, I have no problem with the conclusion. There are confounding factors that are unavoidable and this means that signal detection will be tough. I don't know about the numbers exactly, but economic losses is the last place where any level of certainty would be reached.

    I also wonder what methods and level of certainty or significance insurance companies have used when determining rates for other kinds of risks. For example, when did health insurance rates first get affected by whether you smoke and what evidence did they use. Since they are paying for policies all the time, I expect that they might use something closer to a preponderance of evidence (greater than 50% likelihood), far lower than most physical scientists use. Or maybe they just do anything to increase their rates. ;)

    Lastly, while extreme events are by their nature rare, say when compared with measurements of temperature or changes in the seasons, they aren't all as rare as typhoons/hurricanes. Furthermore, sometimes they correlate. For example a flood in one place could be connected with a drought in a neighboring area. Plenty of commentators have said that the Pakistan floods and the Russian heat wave were due to the same meteorological event. The point being that different kinds of extreme events may be combinable when defining a data set for analysis, which could make detection come much sooner.

  9. -8-Dean

    Some re/insurance companies have already decided that GHGs are impacting extremes. So they need a confidence level of ~0.0 (and you may be hinting at ;-)

  10. I see no reason to waste time disputing the conclusions of this particular paper. Instead, what's important is that the pathway is being explored. Financial risk, climate change, and the overall matter of time discounting in the financial system are going to come together increasingly and I think this--and other explorations--are really important. Well done.


  11. IRT Romm's rationalization to dodge debate:
    He is full of bs and knows it.
    Considering how vile and defamatory and bloviating Romm is, his cowardice is particularly sweet with irony.
    The inevitable comparison (unfavorably) of Romm to Sir Robin comes to mind.

    As to 'rarity', how is a season in one storm basin with an average of ~10, and often double that, 'rare'?
    I believe the message is clear, for those who wish to pay attention: Once again the promised doom of global warming is failing to appear.

  12. Roger: I posted a comment at Romm's place pointing out that it was elementary that slow trends in rare events in noisy signals take a long time to distinguish. It was, of course, deleted.

    My question is: do you think Romm can't follow the Maths? The concept would be taught in High School statistics, so surely within his grasp, but on face value his response suggests his Maths is pretty limited.

    Or is he so lost in the political message that he will attack, even knowing he is wrong?

    It's not like the paper suggests anything about underlying global warming, so its basic premise is not antagonistic to the warmers. So why then does he feel the need to say stupid things in order to counter it?

  13. -12-Mark

    You are correct that the paper is about math and checking our analysis is pretty straightforward. Romm could easily do this if he wanted, but I have learned that substantive engagement is not on his agenda.

    An irony of course, which I suppose Romm doesn't get, is that our paper starts by taking as given the projections that he shows at the top of his "critique".

    Given the straightforward and easily replicable nature of this paper, I do expect that it will be pretty influential in many places, even if the Center for American Progress is not among them ;-)

    Romm's protests about the paper say much more about him than our work, and that is just fine by me.

  14. In the 2009 US Global Change Research Program report 'Global Climate Change Impacts on the US', the summary claims
    "Climate-related changes have already been observed globally and in the United States. These include increases in air and water temperatures, reduced frost days, increased frequency and intensity of heavy downpours, a rise in sea level, and reduced snow cover, glaciers, permafrost, and sea ice." I happen to be working with the NADP-NTN data which has daily rainfall stats at over 200 US sites - ~160 of these have over 30 years of records. Out of curiosity I looked for any indication of "increased frequency and intensity of heavy downpours". There is absolutely none.